AI Is Decompiling Every Game | TheStandup

Summary

The episode explores how AI-assisted decompilation could make it easier to reproduce, modify, and combine games. The hosts distinguish between two kinds of mashups:

• Running games side by side: Each game keeps its own systems while data passes between them. These mashups can be funny, but physics, combat, and animation often behave awkwardly.
• Merging game systems: A game is decompiled so one of its systems—such as a climbing or combat mechanic—can be brought into another game. This could be more useful, though it also raises technical and legal questions.

The panel sees these projects as exciting demonstrations, but not necessarily as games people will want to play for long. AI may speed up the painstaking process of matching compiled code to source code, yet the resulting code can be difficult to read and may need substantial human cleanup.

They also debate what decompilation means for preservation. Keeping a game’s original machine code may preserve the artifact more faithfully. Decompiled source can help people study, modify, or port a game, but poor-quality output may not explain how its original developers built it. The retro community’s concerns are therefore not simply about using AI, but about whether the results are accurate or useful.

On copyright and the games industry, Casey argues that making a close copy may still create legal problems, even if it was produced through a “clean-room” process. He contrasts games—where piracy has long existed and publishers depend partly on customer relationships—with software such as Photoshop, where a free clone could threaten a company’s revenue. The group also worries that powerful companies could use AI to absorb competitors’ ideas, and that restricting downloadable software could leave users dependent on online services.

The discussion ends on the economic consequences of free software: if people are not paid, fewer may be able to do the work. AI assistance might save time, but the panel questions whether it can replace the human effort required to create and maintain genuinely useful products.

Featured People and References

• Participants: Prime, Teej (referred to as “TJ” in the subtitles), Casey Muratori, and Trash.
• Games and examples discussed: Super Mario, Elden Ring, Halo, Animal Crossing, Twilight Princess, Street Fighter, Mortal Kombat, Spider-Man, Batman: Arkham City, Skate, Call of Duty, Super Mario 64, Banjo-Kazooie/Banjo-Tooie, and Doom.
• Other references: Adobe Photoshop, IDA Pro, Radare, Unity, Unreal, and the Game Developers Conference (GDC). The episode also includes a Sentry sponsor segment.